Running Apple Diagnostics fixes slow internal SSD (how on earth?!?)

[vastunghia]

Dear Mac gurus,

I have a 21.5" 2019 iMac (iMac19,2 Core i5 with Fusion Drive) that was recently upgraded as follows:

1. internal 1TB SATA HDD replaced with 2GBTB SDD (Crucial BX500 3D NAND)
2. internal (32MB?) PCIe NVMe SSD replaced with 2GBTB PCIe NVMe SSD (Crucial M.2 PCIe 4.0 NVMe)
3. 8GB RAM replaced with 64GB RAM

For the avoidance of doubt, I have not merged the two drives into one Fusion Drive.

I've been pretty happy with the upgrade, but I'm bumping into an annoying problem.

Basically, every now and then (typically after leaving the Mac sleep for a while), SATA SDD performance drops to unsustainable levels -- say, from 400+ MB/s read/write to 4-6 MB/s. Of course this makes the Mac totally unusable (takes several minutes to open an application, for instance).

Things I tried so far:

• Activity Monitor: no suspicious disk activity (no disk activity at all actually)
• Booting into Win 10 via Boot Camp partition: the problem is still there (confirmed by running disk benchmarks in Windows, and by the fact that loading Windows take some 10 minutes)
• Booting into an Ubuntu USB stick: same as above
• Safe Mode: same
• Reset PMC / PRAM: no help
• SDD SMART status and self-tests: all green lights

The only thing that (temporarily) fixes the problem is running Apple Diagnostics at boot time.

Diagnostics run and find no problem at all. Then I restart the Mac and... magically, the problem is gone.

Till the next time.

Any ideas? What are Apple Diagnostics doing behind the curtain, that is healing my Mac?

Thanks a lot

Sergio

 

[davegoody]

Could be a drive firmware issue, I know that on some older Macs (such as my old Mac Pro) non-Apple SSDs were problematic. Oh, and it's TB not GB else you don't have much space

 

[vastunghia]

Mmh thanks for the info.

Still wondering what Apple Diagnostics is doing to get it (temporarily) fixed though.

Oh, and it's TB not GB else you don't have much space

Ahahah thanks a lot, edited 1st post.

 

[okkibs]

I am sure the diagnostics isn't doing anything. It's probably some disk rescanning, or power cycling that does it. Can you try powering the Mac off when it happens, so that the SSDs lose power? Maybe that fixes it.

Are you sure you have the BX500? It is the mediocre QLC version that has reduced endurancy, performance and 3 instead of 5 years warranty. If possible, I'd replace that. I've had issues with all QLC SSDs I ever used, at the bare minimum I'd do the MX500 although endurance is similarly extremely low.

Part of the issue is that SATA SSDs are pretty dead at this point, and due to the SATA limits on performance manufacturers try to put in the cheapest possible components they can get away with, because you wouldn't be able to notice it anyways over SATA.

I had two BX class drives (although the older BX100 instead of BX500, 500GB and 1TB) and both failed shortly after end of the 3 year warranty.

Otherwise, for your issue, I'd assume either an issue with that SSD in hardware or in firmware. In any case, I strongly suggest replacing it at least for a warranty replacement with the same model, and if you can return and switch to a different model.

There isn't much on a SSD that you can troubleshoot, firmware upgrades are rare, unlikely and might require windows as well (although bootcamp could work). You could check the smart diagnostic status with smartmontools via homebrew. https://formulae.brew.sh/formula/smartmontools Command is smartctl -a /dev/disk0 or disk1.

 

[vastunghia]

I am sure the diagnostics isn't doing anything. It's probably some disk rescanning, or power cycling that does it. Can you try powering the Mac off when it happens, so that the SSDs lose power? Maybe that fixes it.

Done that. Also reset PMC and PRAM, as noted above. Problem does not get solved.

As weird as it sounds, diagnostics is doing something, maybe not directly -- maybe just triggering some disk reset, heaven knows. But running it fixes the problem 100% of the times, and this is happening a couple of times a week, so..

Are you sure you have the BX500? It is the mediocre QLC version that has reduced endurancy, performance and 3 instead of 5 years warranty. If possible, I'd replace that. I've had issues with all QLC SSDs I ever used, at the bare minimum I'd do the MX500 although endurance is similarly extremely low.

Disk Utility reports CT2000BX500SSD1 Media, so that should be it

Otherwise, for your issue, I'd assume either an issue with that SSD in hardware or in firmware. In any case, I strongly suggest replacing it at least for a warranty replacement with the same model, and if you can return and switch to a different model.

Yes, I'll end up doing that, it is just a pain in the neck, I spent loads of time to configure the Mac (and Windows partition) and I have no time to go through it again with a new drive. Time Machine will help, but...

You could check the smart diagnostic status with smartmontools via homebrew. https://formulae.brew.sh/formula/smartmontools Command is smartctl -a /dev/disk0 or disk1.

Done that, as noted. All clear.

Thanks a lot for your help, much appreciated!

Hope someone with some insights on Apple Diagnostics could figure out what is going on when I run it...

 

[okkibs]

I meant for you to post the smart diagnostic (raw) values here. It might give something away.

 

[Nguyen Duc Hieu]

The Crucial SSDs are notorious for fairly high failing rate.
It's not the TLC NAND chip that cause the failure.
It's more on the bad controller chip and power supplying circuit.

 

[vastunghia]

I meant for you to post the smart diagnostic (raw) values here. It might give something away.

Oops!

 

[vastunghia]

I meant for you to post the smart diagnostic (raw) values here. It might give something away.

Oops! Sorry, didn't get it. Please check attached file.

 

[vastunghia]

The Crucial SSDs are notorious for fairly high failing rate.

Good to know. Which brand would you advice?

 

[Fishrrman]

Perhaps I missed it above, but SOMETHING you DIDN'T try:
Take out the 3rd party RAM, and put the original RAM back in.

 

[vastunghia]

Perhaps I missed it above, but SOMETHING you DIDN'T try:
Take out the 3rd party RAM, and put the original RAM back in.

This is interesting. Do you think that faulty RAM could lead to performance degradation specifically limited to I/O speed on SATA drive only (i.e. leaving apparently untouched everything else, including NVMe performance)?

On iMac19,2 you have to remove motherboard to access RAM slots -- not something I will have the time to do any time soon. Also, not sure this would help...

 

[Publicstaticvoid]

Back in the day, a family member had their 2011 iMac internal HDD replaced with an SSD. I had offered to do it for them, but they went with having it replaced at a repair shop. Afterwards, I was surprised to find that they had been charged for a temperature sensor as well. Internet confirmed at the time that it was a necessary thing to do. Maybe that’s what you are missing in your iMac? The slowdown could be because of „overheating“ due to lack of temperature info, and therefore the slowdown occurs. Running apple diagnostics could be resetting something on the firmware, that makes it restart an internal count or something. Just an educated guess, but hope it helps!

 

[okkibs]

In less than a month of uptime enough data was written to decrease lifetime remaining to 96%, and 6.86TB written, endurance is 720TB for this model. That should correspond to less than 1% of lifetime reduction, yet you're at 4% already. That is somewhat strange but might just be a quirk with Crucial, if that were correct it would reduce endurance from 720TB to 175TB which cannot be right. That would be less than a quarter of the advertised endurance.

I checked with a MX500 I have, 360TB endurance 91.57TB written, 64% remaining even though it should be about 74%. Okay, so the lifetime indicator on Crucial SSDs indeed is not working correctly.

In that case there is nothing unusual about these values. But I am actually surprised to find out that my Crucial has such an extremely low endurance, that's absurdly low really.

I checked real quick and Samsung has a 870 EVO 2TB SSD that has a fairly high endurance, though still lower compared to modern M.2 drives. This just leads me to the conclusion never to buy SATA SSDs ever again. The old 860 PRO 2TB looks much better but isn't sold anymore. I think manufacturers dump all the bad hardware into these SATA SSDs.

 

[vastunghia]

Thanks for your comment @Publicstaticvoid, yes with older iMacs one had to include a temp sensor as well (I have first-hand experience with my 2010 iMac), but I think that with newer models this is not the case anymore. In fact, I can see my SSD SATA drive temperature being reported correctly in my Mac.

Notwithstanding,

Running apple diagnostics could be resetting something on the firmware, that makes it restart an internal count or something.

yes, that is my educated guess as well. To bad apparently it will always remain an educated guess.

Thx!

 

[vastunghia]

I checked with a MX500 I have, 360TB endurance 91.57TB written, 64% remaining even though it should be about 74%. Okay, so the lifetime indicator on Crucial SSDs indeed is not working correctly.

Maybe lifetime reduction is not a linear function of written data? Just pure speculation.

I regret not insisting enough with my local reseller for obtaining a 4TB NVMe drive instead of 2TB SATA SSD + 2TB NVMe. On the other hand, he claimed it is almost impossible to find 4TB MVMe drives that are proven to be compatible with my iMac.

Guess I will have to ask him to replace my SATA SSD. What a PITA.

 

[okkibs]

The Corsair Force Series MP510 4TB should be compatible (at least the two lower capacity ones I have work fine in Macbooks) and the endurance rating and 5 year warranty suggests it's a very solid drive, which is why I bought three at once and all three have been outstanding (only 2 are installed in Macs).

It is almost impossible now to prove compatibility, the same drives can have different firmwares and even completely different hardware. Which is a bit bad for a store doing an official ssd swap, since they're on the hook when there's an issue.

Maybe lifetime reduction is not a linear function of written data? Just pure speculation.

Could very well be, but that isn't good. If this trend continues it would mean my SSD would be at 0% life left despite not having reached the maximum amount of written data yet. On Apple SSDs in Macs for example it is linear and corresponds precisely to the amount of data written, as expected.

 

[Fishrrman]

OP wrote:
"Guess I will have to ask him to replace my SATA SSD. What a PITA."

I have a better idea.
It will probably cost LESS THAN having the repair guy handle things again (unless he works for free).

Get a Samsung t7 "Shield" external SSD:

SAMSUNG T7 Shield 1TB, up to 1050MB/s, USB 3.2 Gen2, Rugged, IP65 Rated, for Photographers, Content Creators and Gaming, Portable External Solid State Drive (MU-PE1T0S/AM, 2022), Black

Samsung T7 Shield Portable SSD delivers high performance on-the-go, not matter the terrain. It has an IP65 rating for water** and dust* resistance, with Dynamic Thermal Guard to manage heat control. Transfer huge files instantly. USB 3.2 Gen 2 and PCIe® NVMe™ achieve soaring sequential read/write...

www.amazon.com

This is a USB3.1 gen2 SSD.

Plug it into one of the USBc ports on the back of the 2019 iMac.
Now, set it up to be your EXTERNAL boot drive.

You'll get read speeds of around 850-900MBps from this.
It may not be quite as fast as the internal SSD, but it will be close.

And I predict you will not be bothered with the "slowness problems" you're experiencing right now.

Cheap.
Quick.
Easy.

 

[wilsonlaidlaw]

I have had 100% success rate with Samsung SSD's back when it was possible to change/upgrade the storage on a Mac.

I recently had the battery changed on my M1 Mac book Air (3 weeks out of warranty!!). I first went to my usual repair man and he showed me a stack of failed M1 Macs, where the Texas Industries power controller chip for the soldered in SSD chips had failed and you could see the resulting charring on the SSD's. He had been waiting for some time for replacement SSD's and the controller chips. He apologised for his quote on the battery but apparently Apple shops fit these for less (£129) than he has to pay for the battery on its own, so he suggested I went to the Apple Centre, which I did. Very efficient - back in 2 days by courier.

 

[vastunghia]

It will probably cost LESS THAN having the repair guy handle things again (unless he works for free).

Thank you @Fishrrman, but I guess the repair guy will replace my drive for free as he sold it to me and upgraded my Mac just 4 months ago.

I will take it to him when it gets itself in the "failed" state -- which is a pretty recognizable state, as all of the sudden it takes 10 minutes to load MacOS / Windows + 10 minutes to do login Pretty sure there will be no doubt that the culprit is the new SATA SDD (as reported by benchmark tools as well).

Speaking of which (the benchmark tools), I just tried a bunch of benchmarks in Windows and I keep getting weird results (see below). They look pretty inconsistent to me.

Also, please note that they were performed when my drive seems to be in a healthy state -- no particular lag observed during normal operations. Put it in another way, BlackmagicDesign's tool yielded these results a few minutes later, in MacOS:

which is in line with my expectations for this drive normal performance (or are my expectations too low?).

Any comments?

Thank you

 

[dushant]

@vastunghia did you manage to get to the bottom of this issue? I have the exact same problem and apple diagnostic is the only thing that gets it going temporarily , but the problem returns within 24 hours.

 

[Significant1]

Did you enable trim with trimforce?

MacBook Pro SSD Trim Enabling

Learn how to enable SSD TRIM on your MacBook Pro for improved performance and longevity of your device. Expert guidance for UK users.

scot-comp.co.uk

 

[dushant]

I enabled TRIM using terminal command. It ran well for 24 hours and then slowed down again. Read speeds are finw but write speeds are really bad.

 

[Significant1]

I enabled TRIM using terminal command. It ran well for 24 hours and then slowed down again. Read speeds are finw but write speeds are really bad.

If trim wasn't enabled from the start, I believe only zeroing out the drive (secure erase) can make a clean trim start.
"Regardless of operating system, the drive can detect when the computer writes all zeros to a block, and de-allocate (trim) that block instead of recording the block of zeros. If reading a de-allocated block always returns zeros, this shortcut is transparent to the user, except for faster writing (and reading) of all-zero blocks, in addition to the usual benefit of faster writing into unused areas. Operating systems do not write all zeros to "wipe" files or free space, but some utilities do."

Trim (computing) - Wikipedia

en.wikipedia.org

My knowledge about trim might be/likely is outdated, especial macOS support, so I suggest you do your own investigation.